This document contains the report on the workshop on HCI on the Web. This workshop was held during the Third International World-Wide Web Conference.
The topic of this workshop was "HCI on the Web". Three main subjects had been identified by the workshop organisers:
- World-wide HCI: what do we need?
- The HCI aspects of Web pages: Needs and recommendations
- What changes can be proposed to HTTP and HTML to better support HCI
As usual in good workshops the discussion flowed around a number of aspects related to those two issues. Instead of presenting the discussion in chronological order these aspects will be presented below.
This document contains information about the participants of the workshop, and about all the topics discussed:
- About HCI information resources
- About HCI on the Web
- About WWW as a concept
- About bidirectional links
- About Style Guides & Guidelines for WWW
- About HCI suggestions for HTML
- About style sheets
- About Forms
- About Language
- About adaption to users
- About scrollbars vs. single pages
- About HCI issues of browsers
- About design of URL's
- About usable pages
The following people (sorted alphabetically) participated in the conference:
- Lon Barfield, (email@example.com), General Design.
- Design background. Interested in the design of Web pages.
- Mikael Ericsson (firstname.lastname@example.org), Department of Computer & Information Science, Linköping University, Sweden.
- Maintains a list of HCI Resources, and a list op HCI Thesis In Progress. Also interested in using WWW for education in Sweden.
- Hans de Graaff, Delft University of Technology.
- Maintains HCI Index. Interested in navigation through larger document spaces. Currently working on a virtual conference.
- Nahum Gershon (email@example.com), Mitre.
- Visualization background, puts much emphasis on the user.
- Inke Kolb, (firstname.lastname@example.org), GMD.
- Webmaster at GMD. Currently also webmaster for European Software Institute in Spain. Often too many technical problems.
- Lanny Lampl (email@example.com), Levi Strauss & Co.
- Wants to use WWW for sharing information internal to company. Different perspectives for people, such as geography, subject, etc.
- David Mischel (firstname.lastname@example.org), Genentech, Inc.
- Administrates Genentech web services.
- Steven Pemberton (Steven.Pemberton@cwi.nl), CWI.
- Workshop organiser. Interested in HCI in general. Currently researching electronic books. Member of WWW Consortium.
- Joshua Cooper Ramo (email@example.com), Time-Warner Athena project.
- Interested in broader bandwidth applications and future generations of HTML.
- Kirsten Reimann, EDS, Russelsheim.
- Hassan Schroeder (firstname.lastname@example.org), Sun Microsystems.
- Webmaster of Sun's website. Curious about ideas on offering information to people. Interested on different perspectives of different people.
- Martin Senger (email@example.com), Department of Molecular Biophysics, German Cancer Research Institute, Heidelberg.
- No particular interest in HCI. Interested in guidelines on making useable documents.
- Jon S. von Tetzchner (firstname.lastname@example.org), Telenor Research.
- Has written several Web tools and is currently working on a new browser called MultiTorg Opera.
- Ramon Wartala (email@example.com), DKFZ.
- Webmaster for medical physics department. Wants to create better Web Pages.
Hans and Mikael both maintain indices of resources available on WWW. They briefly describe how their indices came into existence, and what the particular focus of each index is.
Some discussion about the user view starts. What do the users really want out of such resources? It turns out to be very hard to get comments from users. Specific forms for feedback might help here. Another suggestion is to gather a number of users and ask what they need and want. This might in fact be done at the CHI conference.
The current focus is mostly on researchers and students. Additional resources aimed specifically at practitioners would be useful. One example of such a practitioner-oriented resource is Raghu Kolli's UIWORLD.
Examples of what workshop participants would like to see are examples of good design, or an interactive tutorial, including examples. In particular this would be useful for Web pages: which pages are effective and well-designed, and which aren't.
Another problem identified with the current indices is the unstructured format they have. Mostly they are collections of links, grouped together according to some criteria. However, this isn't always the best way to group links. For instance, additional structure added by presenting HCI as a field might be a more coherent backdrop for the collection of links.
The fact that there are a number of different HCI Resources, each with a slightly different focus, makes it hard to look through these pages and find items. It might be more useful to cluster all resources globally, or at least create some global search mechanism. A single top-level page for all resources might also be useful.
A suggestion specific to one particular HCI Resource was to add additional searching capabilities to the HCIBIB database, which would allow searching on references also (e.g. "give me all articles which refer to this article").
Steven brings forward that SIGCHI would like to promote HCI on the Web. Several possibilities were brought forth during the discussion. For instance, a good overview of HCI, the current research issues in HCI, and who is working on them. Also, it might be beneficial to bring together a group of people with interest in helping out or answering questions about HCI. This could be somewhat consultancy-oriented, with added resume's and examples of previous design work.
Along the same lines it might be useful to construct a case-based database which would include examples of good and bad design. By updating this database frequently this would become an evolving information space for good design.
Nahum argues that WWW currently has many conceptual problems. The "Lost in hyperspace" problem is well known. Another problem is that the smallest unit of information is a complete document. He proposes a more personal information space build from existing elements.
He also indicates that this is a new medium. There are new possibilities which we don't even know yet. Computer Scientists probably won't explore the possibilities fully, we need other people work work on this.
Lanny brings up the subject of using bi-directional links. (Each link can also be followed back). This concept was already mentioned by Ted Nelson for his Xanadu project. Also, Hyper-G has some possibilities for this, as it stores all links separate from the documents themselves.
There was no consensus on this issue in the workshop. Critics argued that it would only be useful for providers, not users, it will probably introduce even more chaos, and it might be a ethical problem to know how links to certain pages. In some countries, for instance Norway, this would even be illegal because of privacy protection.
The proponents argued that bi-directional links can be beneficial in getting a better overview of the structure of the Web. Using such links it would also be possible to follow links back to otherwise never uncovered clusters of relevant information. Finally, it would give some measure as to how many people link to a certain page, which might be some kind of justification for creating it, even though this justification might be better based upon actual access.
The people in the workshop agreed that Style Guides or Guidelines are a very useful thing to aid designers. Problems with style guides is that they either have a very narrow focus or are very generic. Also, they are mostly text-oriented.
New guidelines should incorporate the how to write documents for different audiences, and should distinguish between different types of documents. Also included should be examples. What currently isn't clear is how these standards or guidelines should be related to other standards now being used in the computing and information field.
One of Steven's reasons for organizing this workshop is that he is part of the WWW Consortium. He would like to see some discussion on possible extensions to HTML and HTTP, which would allow for better and more usable Web pages.
As an example, Steven brings forward the current ismap (clickable map) construct. It's main usability problem is that it doesn't feature any feedback, either when moving around over it, or when clicked on the map. It was quickly recognized that most of these problems stem from WWW's roots as a text-only system. One solution would be to make graphics and text more equal. However, these problems can be partly solved by designing better graphical maps, with clear visual cues.
Steven notes that these conclusions certainly are true, but that we need to come up with technical solutions to such problems, because this will improve our chances of getting such changes approved by the WWW Consortium. For instance, should the changes for ismap be made to HTML, offering the coordinates through the page, or through HTTP?
We would like to see some new extensions. Currently HTML only knows about the verb "go", and other things (such as animation) are needed. Such an extension should be a generic mechanism with some things build in, and the possibility to extend it. It's not clear whether such extensions could be provided within HTML.
One system which has already implemented several such features is HotJava. For instance, one of its demos features a clickable map with feedback. One problem with HotJava is that it isn't really an architectural solution. There is just one extension <app ...>. This might make it harder to use, although it could be argued that more effort is always required for more powerful things.
Finally, Lon has the suggestion for simple graphics included in HTML, such as line and circle.
Lon brought forward the topic of Style Sheets. Such sheets would reside at the browser site, and determine how to display a certain HTML page. HCI could give examples of better page presentations this way. The main problem with such an approach is that it isn't clear how should decide how a page should look. Currently there still is a conflict between style and content.
An interesting extension would be to add the user context to such a Style Sheet, and make a different presentation for novices and experts, or for people wit slow and fast network connections.
Lanny would like to see state being preserved in forms. For instance, dynamically changing pop-up menu's, comment pages which preserve the correct topic, etc.
With respect to forms Steven points to the FRS (Form Response System). This provides an architectural solution to many forms problems.
Inke wonders about the problems with language. For instance, in Germany she would often have to maintain double trees of documents in German and English. Nahum points out the Hyper-G's meta-data also contains a language-tag which could be used.
Hassan wonders about using the IP address for selecting a proper language or a closer server. Most people seem to think the first isn't a good idea.
David raises the topic of user adaption. For instance, performance of network connections could be taken into account. Currently, this is done in HTML pages (e.g. by specifying a low-grade image to be loaded first, or by presenting a page without graphics separately. The fact that performance changes dynamically will complicate possible solutions.
Some of these issues could be resolved at the browser level, but servers also need to take hints on this and act accordingly.
One possible solution at the browser site would be the use of Style Sheets to indicate or select the behaviour wanted.
This subject turned out to trigger a quite lively debate about which was best. In the end several different positions were recognized, each having equal value.
Those who like scrollbars have the following arguments. Many small pages are much harder to print. Also, a large collection of small pages makes it easier to get lost in hyperspace. There should also be some way to print the information in a collection of web pages, but this is currently usually only possible when information is presented in one large document. Some documents require a strong narrative line, which a larger document can enforce.
Those who dislike scrollbars have the following arguments. Large documents break the webiness of documents, which is centered about having small pieces of information. Another problem with large documents is that links within documents can often be confusing, and browsers don't always handle this well. Scrollbars also seem to impose a hurdle for many people, according to a study done at Sun. (Note that I couldn't actually find this conclusion on this page.) A study at Time-Warner showed that people rather flip the page instead of using scrollbars.
In conclusion we can say that there are different types of documents, such as full documents, maps, indices, information sheets. Each type of document targets a particular audience in particular way, and with some documents scrollbars do work well, and with some they don't.
A study for electronic publication of the SIGCHI bulletin showed that moving some physical properties of books and magazines (such as size, skimming)to the Web publications will certainly help. For instance, the prototype of the SIGCHI bulletin shows all headings first in each page.
Several suggestions for improvements to WWW browsers came up. For instance, temporary bookmarks (for one session only) would make it easier to distinguish between more permanently interesting locations and those only needed in a single session. It would keep the permanent bookmarks list from getting cluttered.
Using multiple screens usually isn't very hard with most browsers, but for some reason not many people use this feature, and it might be interesting to find out why.
The "shopping basket" principle currently employed by many shops on WWW might be made more generic and architectural, which in turn could lead to support for such an architecture in the browser.
The URL's themselves were also part of some HCI discussion. Currently URL's are often intimidating because of there orientation towards the technology used. (cf. with bang-style e-mail addresses) During the workshop some ideas for better design were mentioned, such as simply having a site and keyword, which should be enough.
Another idea might be to change "http" to something more descriptive for the majority of people using it, such as "www" or "web". For instance, web:Steven.Pemberton@cwi.nl might be easier to remember and use. A solution which can be implemented using current URL's would be to always just use a keyword or subject as the only component after the site-name, such as web://some.site/roses.
Something else to take into account when constructing new URL's is to make sure they can be handed out on their own. This introduces some problems, such as the unix pathnames, which can be very short and nondescriptive. However, using longer, and much more descriptive pathnames makes the URL very long again.
Finally, we don't always need URL's. If a user wants to know where a certain links leads to, a brief description of that page (possibly the title) would be much more descriptive.
The current focus on the Web is on Cool pages. Usability hardly ever comes into play. Maybe a SIGCHI award for usable and well-designed pages would be a good initiative to place some focus on the latter aspects.
All the participants enjoyed the workshop. Due to the discussion on many different topics everybody could find something valuable in it. Many issues related to HCI and WWW have been brought forward during the workshop. Even though not many concrete proposals or solutions have come out of it the workshop was still a success. The workshop will be continued at CHI'95, where the current issues can be brought forward again. Possibly, during that workshop, some proposals can be made.